Virtual Machines, VPS and Instances: The Differences That Matter

Virtual machine, VPS and instance all describe a slice of a physical server presented to you as if it were a computer of its own. The technology underneath is similar. The words carry different assumptions, and those assumptions decide how you should treat the thing.

Virtual machine: the technical term

A virtual machine is the neutral description. A hypervisor divides one physical machine into several isolated guests, each with its own kernel, filesystem and network interfaces. It says nothing about who runs it or how it is billed.

VPS: the hosting tradition

Virtual private server comes from the hosting world. It implies a machine you rent for a period, usually monthly, that stays where it is and keeps its address. You care for it: you patch it, you back it up, and you expect it to still be there next month.

The VPS model rewards stability. Pricing is predictable, the machine has a name, and rebuilding it is an event rather than a routine.

Instance: the cloud tradition

Instance comes from cloud platforms and implies something more disposable. An instance is launched from an image, may be billed by the hour or second, and is expected to be replaceable. If it misbehaves you terminate it and launch another.

This model rewards automation. It assumes configuration lives in an image or a script rather than in the accumulated history of one long-lived machine.

Why the distinction has practical consequences

• Billing shape. Monthly renting favours steady workloads. Hourly billing favours bursty ones that can be switched off.
• Attachment. If you treat an instance like a pet VPS, you will hesitate to replace it and accumulate undocumented changes.
• Backups. A VPS mindset leans on backups of one machine. An instance mindset leans on being able to rebuild from an image.
• Addresses. Long-lived machines tend to keep an address. Disposable ones need that address decoupled, or DNS breaks each time you replace one.

Which the providers lean towards

Providers sit at different points on this spectrum, and their catalogue language shows it. DigitalOcean counts droplets, which behave like tidy VPS units. AWS configurations are usually described by vCPU capacity, reflecting an instance-shaped world where you decide the size at launch. Hetzner listings count servers, again a VPS framing.

None of this restricts what you can do. You can run disposable instances on a VPS-style provider and keep a pet server on a cloud platform. The framing simply tells you which habits the platform makes easy.

Choosing a mental model

Pick the model that matches your workload rather than the word the provider uses. A single always-on application server is a pet whatever it is called, and deserves backups and documented configuration. A batch worker that runs for two hours a night is cattle, and deserves an image and a script.

Most trouble comes from mixing the two unintentionally: treating something as disposable while storing the only copy of something on it. Deciding explicitly which one a machine is, and writing that down, prevents most of it. The related question of how big to make it is covered in how to size your first cloud server.